基于Sentinel-2A时序数据和面向对象决策树方法的植被识别

发布时间：2018-07-05 12:50

本文选题：Sentinel-A + 时序数据　；参考：《地理与地理信息科学》2017年05期

【摘要】：Sentinel-2A数据具有较高的空间分辨率和时间分辨率,克服了以往时序数据难以获取或空间分辨率低的问题。该文以山西省吕梁市陈家湾流域为研究区,基于Sentinel-2A时序数据,根据归一化植被指数(NDVI)时序曲线特征和光谱特征,构建基于面向对象决策树方法的分层分类模型,成功提取了陈家湾流域的植被信息,分类总体精度达到89.7%,Kappa系数为0.87。基于面向对象决策树方法的多时相分类结果与单时相分类结果相比,可以有效改善波谱特征相近和受地形影响较大地物的区分,减少混分现象;基于Sentinel-2A时序数据和面向对象决策树分类方法能够有效提高植被分类的精度。
[Abstract]:The Sentinel-2A data has high spatial resolution and time resolution, which overcomes the problem that the time series data are difficult to obtain or the spatial resolution is low. Taking Chenjiawan Basin of Luliang City, Shanxi Province as the study area, based on Sentinel-2A time series data, based on the characteristics of normalized vegetation index (NDVI) time series curve and spectral features, a hierarchical classification model based on object-oriented decision tree method is constructed. The vegetation information of Chenjiawan watershed was extracted successfully, and the overall classification accuracy was 89.7Kappa coefficient 0.87. Compared with the results of single phase classification, the multi-phase classification based on object-oriented decision tree method can effectively improve the classification of ground objects with similar spectral characteristics and affected by topography, and reduce the mixing phenomenon. Based on Sentinel-2A time series data and object-oriented decision tree classification method, the accuracy of vegetation classification can be improved effectively.
【作者单位】：中国科学院遥感与数字地球研究所/遥感科学国家重点实验室;中国科学院大学;
【基金】：国家高技术研究发展计划(863)项目(2014AA06A511)
【分类号】：TP751

【相似文献】